SAT is less predictive for females?

September 12, 2012 at 5:39 am 10 comments

I’d not heard this claim before, seen below in an interesting USA Today piece on trying to get more women into STEM fields.  Is it really the case that math SAT scores are not as predictive for females as males?  I found one study about SAT predictive power, but it doesn’t seem to say that SAT is less predictive for women.  I found other pieces complaining about the predictive power for SAT, but I didn’t see anything about the role of gender.

Not to be ignored is the school’s decision in 2007 to make SAT scores optional in admissions. Tichenor says math SAT scores were not accurately predicting the success of its female students. Historically, average math SAT scores for women have been lower than those for men.

Celina Dopart, who graduated this spring from Worcester Polytechnic with a degree in aerospace engineering and is headed to the Massachusetts Institute of Technology this fall for graduate work, says she submitted her scores, but liked the message sent by the test-optional policy.

via Math and science fields battle persistent gender gap –

Entry filed under: Uncategorized. Tags: , , , .

Essay on what it will take to make MOOCs work Does Google get that teachers innovate?

10 Comments Add your own

  • 1. jane prey  |  September 12, 2012 at 9:00 am

    i heard a speaker [think it was Josh Aaronson — sorry if misspelled/remembered] at the NCWIT Summit speak about this and how if you move the gender identification question to the end of the test instead of at the beginning, that women do significantly better.

  • 2. Don Davis (@gnu_don)  |  September 12, 2012 at 9:59 am

    Well, Jane Prey that would explain differences in predictive power if the SAT did prime stereotype threats (Lindberg et al., 2010). Though there are no gender differences in math ability – SAT scores exhibit gender differences (Gallagher et al., 2000) i.e. the SAT has less predictive power for women.

    That being said – the SAT is pretty much a wealth test (Kohn, 2000). And wealth is the greatest indicator of college completion. So in that regard the SAT is a predictor of college completion.

    Gallagher, A.M., De Lisi, R., Holst, McGillicuddy-De Lisi, Morely, Cahalan, Gender Differences in Advanced Mathematical Problem Solving, Journal of Experimental Child Psychology, Volume 75, Issue 3, March 2000, Pages 165-190,

    Lindberg, S. M., Hyde, J. S., Petersen, J. L., & Linn, M. C. (2010). New trends in gender and mathematics performance: A meta-analysis. Psychological Bulletin, 136(6), 1123-1135. doi: 10.1037/a0021276

  • 3. gasstationwithoutpumps  |  September 12, 2012 at 1:47 pm has a lot of studies of the SAT. It has some predictive power beyond socioeconomic status.

    I occasionally blog about the reports. For example, added a question to their study of discrepancies between GPA and SAT (they should have done a regression using the min of the standardized GPA and SAT, which their results suggested may be a better predictor than either alone, and perhaps better than a linear regression with both).
    (released a month ago) includes tables of differential validity by gender and major, but I’ve not read the report yet, so I can’t say whether SAT is less valid for females.

    They do say
    “Overall, the incremental validity of the SAT for predicting cGPA over HSGPA was larger for female (∆ r = 0.11) compared to male students (∆ r = 0.07), and was particularly large for both female and male students majoring in foreign languages, literatures, and linguistics (∆ r = 0.17 and ∆ r = 0.13, respectively). The incremental validity of HSGPA over the SAT was slightly larger for male compared to female students in the overall sample (∆ r = 0.08 versus ∆ r = 0.06), but was particularly large for female students majoring in computer and information science (∆ r = 0.12), and for male students majoring in communications/journalism (∆ r = 0.10).”


    “There were apparent differences in the distribution of female and male students in the different academic majors in this study. For example, one of the larger differences was within engineering/architecture, in which 6% of all female versus 23% of all male students chose to major. With the exceptions of students majoring in computer and information science and foreign languages, literatures, and linguistics, the relationship between SAT and cGPA tended to be stronger for female than male students. This is consistent with previous research on the differential validity of the SAT by gender that shows stronger correlations for female versus male students (Mattern et al., 2008). ”

    It sounds like CS is unusual in having somewhat lower validity of the SAT for females than males. I’ll have to read the report to be sure I’m interpreting the results correctly, not cherry-picking paragraphs that don’t represent the data well.

  • 4. gasstationwithoutpumps  |  September 12, 2012 at 3:16 pm

    The Mattern 2008 study about differential validity of SAT for different genders is this one, I believe:

  • 5. gasstationwithoutpumps  |  September 12, 2012 at 3:50 pm

    If I’m reading Table 12 in correctly, then SAT scores overpredict grades for CS majors, and overpredict by more for men than for women.

    The SAT tends to underpredict grades for women (in all fields, except CS, and even there the overprediction was larger for men than for women). Of course, the same is true of high school GPA for most fields, so the problem may not be with the test, but with male college students generally doing less well than they did in high school. (Differential partying?)

    Of all the fields, CS seems to be the hardest to predict grades for based on both SAT and HSGPA (lowest correlations in Tables 7 and 8). I wonder if this is because there is nothing on the SAT or in the high-school curriculum that corresponds to computer programming, while other fields generally have the first two years of courses much more like a continuation of high school. The correlation between SAT-math and CS grades for women is the lowest number in the table! Of course, the sample size is really too small: only 80 female CS majors from 18 institutions, only the male “social services and public administration” sample is smaller and it also shows anomalies.

  • 6. gasstationwithoutpumps  |  September 13, 2012 at 1:00 am

    I was getting too long-winded here, so I posted in my own blog:

    • 7. Mark Guzdial  |  September 13, 2012 at 1:10 pm

      Thanks! I appreciate the follow-up!

  • 8. Paul Buis  |  September 13, 2012 at 8:45 am

    I just saw stats for a study of Physics majors which showed that the average Math SAT score for majors successful in Calculus 1 was lower than the average Math SAT score for majors who were unsuccessful. On the other hand, average high school GPA was significantly higher in the group that was successful. Some folks around here like to look at high school class rank percentiles rather than raw GPAs as a meaningful predictor of success.

    Our math department has been using its own acheivment test to predict success in math course (and place students into the higest level course they are likely to succeed in). The claim of the SAT is to measure aptitude, not achievement. Aptitude might be a good predictor of long-term success, but is not a good predictor of success in introductory courses. Achievement is influenced by attitude and high-school experiences.

  • 10. Paul Buis  |  September 13, 2012 at 8:50 am

    We’ve got even worse problems with the GRE. I want a test to measure analytical thinking, not quantitative abilities or analytical writing. For non-native English speakers (most common group of applicants to our grad program), the verbal score is essentially meaningless and the analytical writing score tells me more about their writing skill than their analysis skills. For us, requiring the subject GRE for CS scares away too many applicants (or so my collegues fear).


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Trackback this post  |  Subscribe to the comments via RSS Feed

Enter your email address to follow this blog and receive notifications of new posts by email.

Join 9,005 other followers


Recent Posts

Blog Stats

  • 1,878,994 hits
September 2012

CS Teaching Tips

%d bloggers like this: